Template-Based Modeling of Protein-Protein Interfaces

Abstract

Modeling of protein binding sites is important for 3D prediction of protein complexes. Statistical analysis of target-template PSI-BLAST sequence alignments was performed for 329 two-chain target protein complexes selected from DOCKGROUND database. For 214 complexes (∼65 %) the alignments contained all interface residues (full interface coverage or FIC alignments) for both complex monomers and 101 (∼30 %) complexes had FIC alignments for one of the monomers. The FIC alignments were observed even in the case of poor alignments where only a small portion of the target sequence (as low as 40%) was aligned to template sequence with low alignment identity (40%) alignments, whereas for the low-identity alignments the situation is opposite. Homology models were built based on the FIC alignments with target sequence coverage < 60 %. The results showed that one third of the target sequences with such short FIC alignments produced models with interface RMSD (i-RMSD) < 5 Å, suitable for low-resolution ab initio docking. The proteins with i-RMSD < 5 Å had domain structure, whereas models with 5 Å < i-RMSD < 8 Å (accuracy suitable for structure-alignment methods) were generated for single-domain proteins as well. The results provide guidelines for building 3D protein models for docking studies.